Automated Model-Based Clinical Summarization of Key Patient Data

SHARPLeaders: Dean Sittig, Adam Wright

Most clinical information systems are organized around data types (e.g. lab results, medications, problems, vital signs and notes); however, this organization does not usually match the cognitive patterns of clinicians, which tend to be problem-oriented and cut across data types. Project 3 is dedicated to developing methodologies for modeling and summarizing complex chronically-ill patients’ electronic health records, which will be enhanced with context-appropriate, evidence-based recommendations that improve clinician decision-making under information overload and time pressure. Creating such summaries is challenging, and depends on both a deep understanding of clinician cognitive processes and accurate models of clinical knowledge and practice. We will use the Rapid Assessment Process (RAP), a modification of traditional ethnography, to understand clinician summarization needs and to develop clinical requirements . After this portion of the project is well-underway, we will design automated methods of creating accurate, succinct, condition-dependent and independent computer-generated summaries of complex, chronically-ill patients with the ultimate goal of improving patient safety, clinician efficiency and satisfaction, and reduce the cost of care.

Products

  • MAPLE Knowledge Base: A validated knowledge base that can be used to infer problems from medications, laboratory results, billing data, procedures and vital signs. The knowledge base is available at http://jamia.bmj.com/content/18/6/859/suppl/DC1 and is described in a paper cited below.

  • Problem-Medication Linkage Knowledge Base: An ontology-based knowledge base containing nearly 34 million distinct problem-medication pairs by:

1) using the “may_treat” relationship within NDF-RT, mapping the medications and problems to RxNorm and SNOMED-CT, and
2) inferring additional relationships using the “ingredient_of” and “isa” relationships between similar medications in RxNorm and derivative problems in SNOMED-CT.

An early version of this work with 7 million problem-medication pairs was described in an AMIA proceedings paper, cited below, and a substantially revised version is listed below as: A method and knowledge base for automated inference of patient problems from structured data in an electronic medical record. The full KB (34 million pairs) is included as an Appendix in this manuscript.

  • MedEx: Medication information is one of the most important types of clinical data in electronic medical records. It is critical for healthcare safety and quality, as well as for clinical research that uses electronic medical record data. However, medication data are often recorded in clinical notes as free-text. As such, they are not accessible to other computerized applications that rely on coded data. We describe a new natural language processing system (MedEx), which extracts medication information from clinical notes.

The code for this project can be found at: http://code.google.com/p/medex-uima

Demo

Demonstration of Patient Summarizer within the SMART App Platform

Publications

SHARPC Logo